Advice on when to apply for CompLing PhD and writing samples

[ergative]

Hi everyone! I am currently in a business analytics master's degree program, but after a lot of thought over the past few months, I have decided that I want to pursue my original interest - computational linguistics/natural language processing - at the PhD level. I've already compiled a list of programs I'm interested in so far based on the research they are doing.

However almost all of them require a writing sample as part of the application. This is a problem because my current program has 0 research whatsoever, I have no writing samples at all I could submit. I no longer have the research paper I wrote in my linguistics BA as my senior exit requirement, though it's a phonology paper not related at all to computational linguistics, and I also wrote it 8 years ago, so probably not a good sample anyway. I would have to write something from scratch. Looking for general advice here - how would you approach trying to find an issue to research that can be easily written up in only 10-15 pages? All the ideas I have seem too big in scope to be covered in such a short paper. I've heard that "compare and contrast" papers are frowned upon, but what original research could I do that's that short??

Because of this issue with the writing sample, I am concerned about being able to apply for this application cycle, as the earliest deadline I have is December 1. I'm not sure I will have the ability to produce a quality paper in this time. The schools would also only see last year's grades on my transcript, they wouldn't be able to evaluate me based on my entire performance in this program - and I'm taking the more advanced classes now, the first year was mostly prerequisites thanks to my BA in linguistics. Also if I wait a year, I could potentially do independent study next semester and work on a small research paper and produce a writing sample that way (a thesis is not an option, sadly), but I'd have to get approval from the department and I haven't discussed it with them yet.

Do you think it would be better to wait until next fall to start applying under these circumstances? I would much prefer to start the PhD next fall, but if I'd have better chances waiting, then I will have to do that. I wish I had made up my mind sooner! ? Any advice at all would be appreciated!

[gdala]

With your business analytics studies, I'm guessing you know how to do at least some machine learning/statistical modelling? If so, I think the most straightforward option would be to find a public dataset (from Kaggle, UCI, etc) that ideally involves language or some form of NLP. Try different types of techniques and models to solve the problems and research the methods you use. Even if you don't come up with a great model, you could still have a quality research paper.

Another idea is to choose a problem in the area you want to research, for instance, problems for computational models for parsing. You could do a literature review and try to implement current state of the art methods and note your results or possibilities for new solutions to the problem.

Neither of these ideas are likely to be groundbreaking papers, but I think they have potential to be decent writing samples in your timeframe.

 

[ergative]

We haven't gone too in depth yet (my current classes are going into more depth), but I am familiar with a few statistical modeling and machine learning techniques already. That sounds like a great idea! I will definitely check out the datasets on Kaggle (or elsewhere) but I wasn't sure if that would be good enough for a writing sample. I'll consider a literature review as well if I don't find a dataset I would want to explore. I'm not looking for a groundbreaking paper here, I figure that's the point of a PhD - just want something good enough for a writing sample.

I already had one idea for a paper but it sounds kinda dumb to me - I was looking at sentiment analysis in Twitter and noticed everything referenced Twitter's 140 character limit, but they raised the limit to 280 last year. I was curious to see how the raised limit would effect the results of current methods. Is that something that could work for a writing sample or is it kinda a silly question to ask?

[ebs]

Apologies for jumping in really late here, but you may want to consider applying for a master's degree in computational linguistics before applying for a PhD. (Good) PhD programs in CL are very competitive because of the rapidly increasing interest in the field and having a solid field-specific course background as well as having the opportunity to do original research in a professor's lab would really improve your application chances. I'd say to be competitive you need a solid understanding of most deep learning techniques in NLP (MLP, RNN, CNN, LSTM/Bi-LSTM, Autoencoders, etc.)